Advanced Network Approaches for Wireless Environment Branislav JARÁBEK Slovak University of Technology Faculty of Informatics and Information Technologies Ilkovičova 3, 842 16 Bratislava, Slovakia beejay@orangemail.sk Abstract. An ad-hoc network is a special kind of network, where all of the nodes move in time. So the topology of the network changes as the nodes are in the proximity of each other. Ad-hoc networks are generally self-configuring, because of the concept of no stable infrastructure takes a place. There are two different research areas according to wireless networks: Networking for Pervasive Computing, and Wireless Ad-Hoc Networks. This paper focuses on a Service Discovery problematic from the first area describes existing discovery protocols as JINI, UPnP, etc. and routing algorithms in MANETs from the network point of view, which are now-a-days as experimental internet RFCs. 1 Introduction Wireless networks in today meaning are usually set up with a centralized access point for each area of connectivity. The access point has knowledge of all devices in its area and routing to them is done in a table driven manner. Following the trend of mobility, it becomes important to give users the possibility of finding service they need anywhere in the network. Users would like to obtain access to services automatically, without reconfiguring their system. Especially with mobile devices dynamic discovery of services in other than home or corporate network and automatic network configuration will be very useful. An ad-hoc network is a special kind of network, where all of the nodes move in time. So the topology of the network changes as the nodes are in the proximity of each other. Ad-hoc networks are generally self-configuring, because of the concept of no Supervisor: doc. Ing. Margaréta Kotočová, PhD., Institute of Computer Systems and Networks, Faculty of Informatics and Information Technologies STU in Bratislava M. Bieliková (Ed.), IIT.SRC 2005, April 27, 2005, pp. 108-115.
Advanced Network Approaches for Wireless Environment 109 stable infrastructure takes a place. Each node in the network must be able to take care of routing of the data and this is the domain of ad-hoc routing. So the concept of centralized network with pre-defined central routing tables could be applied only with difficulties or with the loss of the advantage of on-the-fly nodes changing. There are two different research areas according to described problems in wireless networks [1]. First is Service Discovery Algorithms, which focuses on service propagation and discovery. Second research area is Mobile Ad-Hoc Network, which maintains design, prototype and evaluates protocols and algorithms to facilitate adoption and use of wireless ad-hoc networks for appropriately selected application areas and facilitate development of standards for such networks. 2 Service discovery Future software services and automated devices will exist in large numbers and will operate in a networked world where they can never be quite sure about the connectivity available, about the other services and devices nearby, or about the state of the network neighborhood a few minutes in the future. In such uncertain environments, individual components will need to discover and maintain awareness of their surroundings and to configure and adapt themselves in response to changing situations. The beginnings of the necessary capabilities for service discovery have emerged over the past few years in the form of service and device discovery protocols, such as Jini Networking Technology, Universal Plug-and-Play, (UPnP) the Service Location Protocol (SLP), Bluetooth service discovery, the Salutation Consortium protocol suite. Despite differences, all of these initiatives aim to allow small collections of commercial devices and services to auto-configure, to cooperate, and to adapt to changes. 2.1 Service Location Protocol (SLP) The Service Location Protocol is developed by the IETF working group called SvrLoc [2]. Focus of this group is to design a vendor-independent standard based on TCP/IP network. SLP architecture consists of three main components: Service Agents are service careers, takes a place somewhere in the area (location), which is advertised with service characteristics. User Agents are client applications, which use services, so they are service discoverers. Directory Agents are collectors of information received from Service Agents and they responds to User Agents on their queries. They could also create different groups of services and service agents and categorize them. The concept is very simple. Each new service must register its service to the Directory Agent. When a user wants a certain service, the User Agent asks Directory Agent for available services in the network. Directory Agent sends appropriate addresses and characteristics and user could select and use one. The main goal before this concept could be reliable is how to find a Directory Agent. There are three different methods defined: static SLP agent obtains an
110 Branislav Jarábek address to Discovery Agent via DHCP while connecting into the network, active User and Service Agents request to the multicast group and listening Directory Agent replies directly via unicast, passive Directory Agent periodically sends multicast advertisements and User and Service Agents listen to them, and after receiving that multicast are able to communicate via unicast. There is also a concept, which omits Directory Agent. If there is no DA (for example small home networks) UA repeatedly send multicast requests for a service until one or more listening SAs send out their response via unicast. 2.2 JINI Jini technology [3] is based on the Java language. Jini creates an issue of how devices connect with each other in order to form a simple ad-hoc network (a Jini community ) and how these devices provide services to other devices in this network. Jini consists of architecture and programming model. Each Jini device must have running a Java Virtual Machine. The architecture principles are similar to that of SLP. Devices and applications register within Jini network through a Discovery and Join process. Application also places itself into the Lookup Table on a lookup server (always needed). Besides pointers stores Lookup Table also Java-based program code for these services. This means that services may upload device drivers, an interface, and other programs that help user to access the service. When client wants to utilize the service, the object code is downloaded from the Lookup Table to the JVM of the client. This code mobility replaces the necessity of pre-installing drivers on the client. 2.3 Bluetooth SDP Bluetooth is technology created for wireless short range transmission. Bluetooth protocol stack contains also a definition of the Service Discovery Protocol (SDP). This protocol is designed to categorize and to advertise the services provided by each node. But only advertise. It is specifically developed for this short range environment. Discovering could be done via searching for service by service attributes, searching for services by service type and browsing without any service characteristics. SDP does not cover accessing services. Any other mechanism as selection and accessing are out of scope of SDP. But there are co-existence patterns with other service discovery protocols, but it does not require them. 2.4 UPnP UPnP Device Architecture [4] is more than just simple extension of the plug and play peripheral model. Universal means that no device drivers are needed common protocols are used instead. UPnP devices can be implemented using any programming language.
Advanced Network Approaches for Wireless Environment 111 Each device must have a DHCP client to connect to the network and obtain appropriate address. Then five steps take care about device and service handling: 1. Discovery discovery protocol (Simple Service Discovery Protocol SSDP) allows device to advertise its service to control points on the network. Similarly, when a control point is added to the network, the UPnP discovery protocol allows that control point to search for devices of interest. 2. Description has two parts: device and service description. Device description contains vendor specific information. For each service included, the device description lists the service type, name, and other access URLs. Service description includes a list of actions with parameters, the service responds to, and also list of variables of the state of the service. 3. Control a control point can ask services to invoke actions and receive responses. To do this, a control point sends a suitable control message to the control URL. In response, the service returns any results or errors from the action. Action could change state of the variables that describe internal state events are published to all interested control points (see eventing). 4. Eventing uses standard publisher-subscriber model. 5. Presentation retrieving it is a simple HTTP-based process. 3 Mobile Ad-Hoc Networks (MANETs) There are plenty of routing algorithms suitable for ad-hoc networking problems. Conventional approaches to routing like link state or distance vector are well tested. But the main problem is that they are made for a network with static topology. They will be working well for MANETs with low dynamic parameters, but they could be devastating for larger MANETs because of their dependency on periodicity of control messages and potential large number of destination hosts (network nodes). Due to mobile devices resources (battery power, bandwidth, CPU for example PDAs) and wireless environment constraints (e.g. not bi-directional) both link-state and distance vector wastes resources trying to maintain routes to all reachable destinations and must be modified to achieve new performance issues. [5, 6]. MANET IETF working group will develop two standard track routing protocols for: Reactive MANET Protocol (RMP) and Proactive MANET Protocol (PMP). From the most common candidates for reactive protocols DSR (Dynamic Source Routing) and AODV (Ad-hoc On-demand Distance Vector, experimental RFC) routing and from proactive approach OLSR (Optimized Link State Routing) and Topology Broadcast (Dissemination) based on Reverse Path Forwarding (RFC Draft) are described below. Reactive Ad-Hoc Routing Algorithms use on-demand routing, so the routes are only requested when they are needed. This approach stores only active routes in memory and does not waste bandwidth maintaining information about unused routes.
112 Branislav Jarábek Main disadvantage is the latency before first packet is sent caused by the route discovery. Proactive Ad-Hoc Routing Algorithms (table-driven) maintain routes to all nodes in the network, so they heavily rely according to tables. Difference to reactive is that there is no need to construct new route before transmitting but only fetch the next hop node from routing table. 3.1 Dynamic Source Routing (DSR) Main concepts of DSR [7] are route discovery, route caching and route maintenance. Routing by this method means that the path is discovered. The first data (this means that no data has been sent to that destination before) initialize the route discovery from the source to the destination. Route discovery: Can be done in two ways active or passive. A source node can flood the network with a route request (RREQ) for a node it needs a route to. Its neighbors propagate this RREQ in the same manner until it reaches the destination node or a node that already has a route to the destination. Each hop of the RREQ is recorder within the RREQ message. The second way for a node to discover a route is by caching RREPs that it happens to pass on or overhear from other nodes. Route caching: Upon arrival of the RREQ at the destination node or at a node that already has a route for the destination, the inverse route is cached. In the case of a node that already has a route, the existing route is appended. A route reply (RREP) is then sent back the same way the RREQ traveled. The RREP contains a list of all the intermediate nodes between source and destination. On the way back, the RREP is cached by all intermediate nodes it passes on the way. When the RREP reaches back to the source, the source can send the data and cache the new route for future use. Route maintenance: Whenever a node discovers that the link to one of its neighbor nodes has failed, it will send a route error (RERR) packet to the source of any cached routes that use the failed link. This RERR will propagate through all intermediate nodes that also have cached this route. Upon receipt of a RERR a node will update their caches accordingly and purge the now failed route. In the case of nodes using promiscuous listening, the node should note that all RERRs might not be heard (the node may have moved) and the route should be treated differently than route where the node is an intermediate node. 3.2 Ad-hoc On-demand Distance Vector Routing (AODV) As the name applies, this protocol adds specific on-demand modification to the distance vector routing algorithm [8]. The reason to this is to acquire low bandwidth cost as DSR has with lower time to discover a route. The idea behind is to use only local broadcast and also modified discovery algorithm. In route discovery AODV uses RREQ packets as DSR in the same meaning packet is broadcasted into the network. The difference is that the hosts on the way of RREQ packet to the destination maintain back pointers from where they received the
Advanced Network Approaches for Wireless Environment 113 packet. This creates reverse path from the destination to the source. When the packet reaches the destination (or a host whit path to the destination), the destination sends a RREP packed back, which validates reverse path and sets up the forward path. All hosts, which maintain back pointers and do not receive RREP packet in specified time, will de-allocate the pointer. This approach guarantees that the only one route is in use. So this is the difference to DSR algorithm first builds up a tree of possible paths from source to destination and then validates only one of these routes. The only one shortest path can be selected when intermediate nodes uses the sequence numbers of the RREQ and RREP packets (in opposite to DSR which can multiple routes). So this algorithm is better to use in larger networks (path is not stored in packets, only one route from the tree of paths is used). Route maintenance: If the source node moves, nodes on the path will send a special RREP. If next node fails to communicate, the node will send information upstream the path (RREP failure packet). All nodes will remove appropriate information from their route caches. Local connectivity management: AODV uses local HELLO messages which transmit to the immediate neighbors. These messages detect link failures in active routes as described before. 3.3 The Optimized Link State Routing Protocol (OLSR) OLSR [9] is an optimization of a link state protocol for mobile ad-hoc networks. OLSR uses Multipoint relays (MPRs) and MPR Selectors to reduce the bandwidth cost of maintaining routes by limiting the broadcast of link-state updates. So each node selects a subset of its neighbors, as MPRs and these nodes are the only ones that forward packets from this node. The MPRs record that another node has selected it as MPR so node is MPR Selector for MPRs. OLSR uses local broadcast of HELLO messages, which are no forwarded more than one hop so nodes maintain bi-directional links. Each HELLO message contains a list of neighbors, which this node has received a HELLO message from. This is recorded in a neighbor-table with sequence numbers. If a node receives HELLO messages from all of its neighbors, it has knowledge of all nodes in the distance of two hops. The MPR set should be smaller than the set of neighbors, to take advantage of relaying through another host. The MPR set is included in HELLO messages, so the nodes, which take this node as MPR, know to which nodes is this node MPR Selector. The MPR set is recomputed whenever the link to the neighbor node goes down or a new node is added. For calculating the routing table OLSR uses a Topology Control message. This packet is broadcast to all nodes of the network, using MPRs and contains the nodes for which source node is MPR Selector. These packets can be used by the node as the list of last-hop nodes for destination node and are stored in topology table with a sequence number. Constructing of the route table begins with appending of entries of topology table (node, last-hop node) starting from each destination in the network and followed until a node in neighborhood is the last-hop node. So the complete routing table of the
114 Branislav Jarábek network is created and only optimal routes with minimal hop are chosen. When any of the entries in topology or neighbor table change, the routing table has to be recalculated. 3.4 Topology Broadcast (Dissemination) based on Reverse Path Forwarding (TBRPF) TBRPF [10] is pro-active link-state routing protocol based on existing reverse path forwarding algorithms. It has two modes of operation and two main system functions: discovery and routing. The operating modes Full Topology (TBRPF-FT) and Partial Topology (TBRPF- PT) are different in amount of information each node has. As name implies in full topology mode each node has knowledge of all links in network. In partial topology mode, a node has enough knowledge to compute minimal route (min. hop distance) to any other node. Each topology has its advantage in other network: FT mode is suitable in open networks with high scattering and PT mode is better for use in dense networks where the amount of routing information is high and all the information is not needed. Like OLSR, TBRPF sends a Hello message to first-hop neighbors. But the message differs TBRPF sends out only Ids of new and lost neighbors in Hello message, not the complete set. A new node is node that has been recently seen, but two-way link has not been established. A two-way link is in the meaning when the new node responds with a HELLO and receives an update request. This brings decreasing of control data, which is sent over network, and increased frequency of messages could minimize convergence time. Topology update and routing function: Each node keeps a source tree of links, which determines the network topology. A link (a, b) is in the source tree of a node if it is reported by the next hop parent node to be the shortest path to a. The source tree is the basis for a node's reportable sub tree. The report from node b to node a include b's links to all its neighbors and the branch of its source tree rooted at the node that b is the next hop of for a to reach. In FT and PT mode forwarding of update messages is done only for link, which does not contain a leaf node. PT also restricts forwarding so only changes of source tree of the node are reported. In this approach each node creates its source tree and computes route to the destination by computing the minimal-hop path. TBRPF also uses aggregation of control messages to minimalize the total amount of messages sent. 4 Conclusions This paper has cleared research in two major areas of wireless networking of today. The focus has been given to the new service oriented view on the computer network and also routing standards has been considered. Regarding the mobile device s reduced power supply, the radio transmission have to be as low as possible. So on-demand
Advanced Network Approaches for Wireless Environment 115 (reactive) protocols have been introduced in opposition to proactive, as they do not need constant topology updates. It will be possible to aim next work to couple routing techniques with basic discovery services (as simple as for example Bluetooth SDP has) in small networks and measure performance against severity of routing and discovering separately. Acknowledgement: This work has been supported by the Grant Agency of Slovak Republic grant No. VG1/0157/03. References 1. The Advanced Network Technologies Division of National Institute of Standards and Technology, http://w3.antd.nist.gov 2. Service Location Protocol, RFC 2165, June 1997, http://www.ietf.org/rfc/rfc2165.txt 3. AR-Jini Architecture Specification, Sun Microsystems, January 2005, http://www.jini.org/nonav/standards/davis/doc/specs/html/jini-spec.html 4. UPnP Device Architecture version 1.0, Jun 2000, http://www.upnp.org/download/upnpda10_20000613.htm 5. Larson, T., Hedman, N.: Routing protocols in Wireless Ad-Hoc Networks A Simulation Study (Master Thesis), Lulea Tekniska Universitet, ISSN: 1402-1617, 1998. 6. Mobile Ad hoc Networking (MANET): Routing Protocol Performance Issues and Evaluation Considerations, RFC 2501, January 2001, http://www.ietf.org/rfc/rfc2501.txt 7. The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks (DSR), RFC Draft, July 2004, http://www.ietf.org/internet-drafts/draft-ietf-manet-dsr-10.txt 8. Ad hoc On-Demand Distance Vector (AODV) Routing, RFC 3561, July 2003, http://www.ietf.org/rfc/rfc3561.txt 9. Optimized Link State Routing Protocol (OLSR), RFC 3626, October 2003, http://www.ietf.org/rfc/rfc3626.txt 10. Topology Dissemination Based on Reverse-Path Forwarding (TBRPF), RFC 3684, February 2004, http://www.ietf.org/rfc/rfc3684.txt